Implicit conversion from float -23.896848137535812 to int loses precision